EXCEPTIONS ARE NOT NEW TO YOU
Assuming you have completed part 1 of this tutorial, you have seen many references to exceptions, but we have said little about how you can use them. The purpose of this chapter is to instruct you on the use of exceptions, and by the time you complete it, you will have the ability to use exceptions to develop a program with its own error handling ability.
WHY DO WE NEED EXCEPTIONS?
The original charter for the development of Ada included the ability to operate in a real-time environment. You already understand, if you have much programming experience, that if it is possible for an error to surface, it will eventually surface. Many programming languages simply terminate operation if a "fatal error" is detected, but this could be a disaster if the program was controlling a real-time system upon which human lives or safety depended. A 747 in final approach to the airport, or a system used in a hospital operating room would be two examples of systems that simply could not be terminated abruptly because a bad data point was somehow accumulated. The careful application of Ada exceptions will allow the software to gracefully recover from such a situation rather than aborting operation completely.
OUR FIRST EXCEPTION
Example program ------> e_c17_p1.ada
Examine the program named e_c17_p1.ada for our first example program with an exception handler. Ignore lines 18 and 19 for the moment and you will have a program that is not at all unusual, and should pose no problem for you to understand. The program does have a carefully introduced error however, because when we reach a value of 4 for Index in line 14, we will be attempting to divide by zero. Dividing by zero is not allowed in any programming language, because the answer is infinite and therefore undefined. The Ada runtime system will, by definition, cause the exception named Constraint_Error to be raised, which is the Ada way of saying that a divide by zero was attempted. This signals the system to do something about it. (Actually, there are many other ways to get the Constraint_Error exception raised but we will worry about them later.)
The Ada system will search, in a very definite way, for any instructions we have given about this error, and if it finds none, will terminate operation of the program after issuing a message concerning the error. If we have given instructions about what to do with the error, it will execute the instructions and continue operation as we direct it to do. The method of giving the system these instructions is illustrated in lines 18 and 19.
HOW ARE EXCEPTIONS HANDLED?
When any exception is raised, the system immediately looks at the end of the current block or subprogram for the reserved word exception. If it is found, and if the specific exception that was raised is defined there, the instructions associated with that exception are executed, and the subprogram or block is exited.
To define a handler for a specific exception, the reserved word when is used, followed by the name of the exception, and finally the sequence of statements to be executed following the => operator. The sequence of statements can be of arbitrary complexity, but should be kept simple due to the nature of exception handling. In this case, we output a message to the monitor and do nothing else. As many different exceptions as desired can be handled at the end of any block or subprogram by adding additional constructs of the form,
when <exception-name> => instructions;following the single instance of the reserved word exception. We will study examples of multiple exception handlers later in this chapter.
WHAT HAPPENS FOLLOWING EXCEPTION HANDLING?
Following handling of the exception, the program executes an otherwise normal return to the calling program, and the normal sequence of instructions in the calling program is executed. Note that it is impossible to jump back into the subprogram or block in which the exception was raised from the exception handling routine at the end of that block. In this case, because of the logic used, the loop defined in line 10 is terminated early because we essentially jumped out of the loop and the program is ended. If an exception handler or a group of exception handlers is included, it must be the last thing in the block. If normal execution of the block reaches the end of the executable statements by coming to the reserved word exception, the block is terminated normally. You cannot drop into the exception handler at the end of a block. The only way to get to the exception handling code is through raising an exception.
In spite of the additional questions you have at this point, compile and execute this program. Observe the results, and if you do not understand the output, reread the above text until you do, because these fundamental points are essential to understanding the entire topic of exceptions.
LET'S USE SEVERAL EXCEPTIONS
Example program ------> e_c17_p2.ada
Examine the program named e_c17_p2.ada for additional examples of exceptions. This program will answer many of your questions about exceptions.
In the last program we terminated the loop when the exception was raised, but we may desire to continue execution of the program following the exception. The first procedure in this program is logically identical to the last example program except the loop is moved to the calling program. When the divide by zero is detected by the system, which raises the Constraint_Error exception, the exception is handled by the exception handler defined in lines 17 and 18, and the return to the calling program is effected. In this case however, when control returns to the calling program, we are still inside of the loop, and the loop completes normally. This should indicate to you that by careful selection of where you handle exceptions, you can control the overall result. We will see more about this as we continue our study of exceptions.
The logic of the second group of instructions, found in lines 49 through 64, is identical to the logic of the first group as studied in the last paragraph. The only difference is that the procedure has been changed into a block and inserted into the code in an inline fashion. This has been done to illustrate the use of an exception in a block of code, and to illustrate that the exception handler for the block of code is put at the end of that block. After the exception is raised and handled, execution begins at the first statement following the block. Because the block is contained within the loop, the exception is handled within the loop and the loop runs to completion.
MULTIPLE EXCEPTION HANDLERS
Finally, we come to the section of code in lines 66 through 69, consisting of a simple loop calling the procedure New_Divide_Loop. The procedure itself, defined in lines 21 through 40, contains an example of a new operation, the ability to make up our own exception, raise it our self, and handle it with our own exception handler.
Line 22 declares the identifier My_Own_Exception as being the name of an exception, and is defined in much the same way that we would declare a variable. We cannot assign a value to it, but we can raise it anywhere within its defined scope which is the same as the scope of a variable declared at the same place. The exception is automatically initialized to the "not raised" condition by the system.
Beginning in line 34, we define three different exception handlers, which will cover any exceptions raised anywhere within this procedure. The first two are named exception handlers but the third handler uses the reserved word others to indicate that it will be used for any exceptions that are not handled by the two named exception handlers. The others clause is optional, but if it is included, it must be last.
RAISING AN EXCEPTION
If we reach line 28 with a value of 4, which we eventually will because of the logic of the calling program, we will detect the divide by zero that would be attempted upon reaching line 31. Instead of letting the system generate the exception named Constraint_Error, we generate our own exception named My_Own_Exception, using the reserved word raise followed by the name of the exception. As soon as we raise this exception, the system jumps to the end of the block, looks for the reserved word exception, which it finds, then looks for the exception handler with the name that was raised. Upon finding it, the statements are executed, resulting in a message being output to the display, and we return to the calling program.
In this case, the system will not raise the exception Constraint_Error, because we are detecting the error before it actually happens. You could raise it yourself by inserting the statement "raise Constraint_Error;" somewhere in this procedure, possibly when the value of Index is equal to 3. It would be a good exercise for you to insert that in the code to see that you can raise one of the system exceptions as well as your own.
Be sure to compile and execute this program to verify proper operation according to this description and to ascertain your understanding of the same.
Note that if an exception occurs, formal parameters of mode out or in out are not updated since a normal return is not accomplished. Partial results will therefore not be returned to the calling program. This is not illustrated here, but is left for the student to investigate if desired.
WHAT ARE THE PREDEFINED EXCEPTIONS?
There are four predefined exceptions which can be raised by the system to indicate a very specific problem. A brief definition follows;
Example program ------> e_c17_p3.ada
Examination of the program named e_c17_p3.ada will reveal what happens if an exception is raised that is not handled by the program. In a word, the program will be terminated, but we need to understand how termination occurs so we can intelligently prevent it.
There is a loop in the main program which calls two procedures successively, Divide_By_Zero and Raise_An_Error. The first procedure is identical to that in the first two example programs and the only exception raised is Constraint_Error, which is handled properly.
The second procedure has its own exception defined, named My_Own_Error which it raises and handles itself in the manner defined previously in this chapter. It also has a divide by zero problem in line 26 that will raise the exception Constraint_Error when Count is equal to 6. Of course, the logic is defined to make this happen and illustrate the error.
PROPAGATION OF EXCEPTIONS
When the exception Constraint_Error is raised at line 26, the system searches for the reserved word exception which it finds in line 31 at the end of the procedure. It then searches for a sequence of statements for Constraint_Error which it does not find. Since an exception handler is not found within the procedure, the exception is propagated to the calling program in such a way that the exception appears to have been raised by the calling statement. In this case it will appear to the logic as if the exception Constraint_Error was raised by the statement in line 39. Once again, the exception rules are applied, and the system searches for an exception section at the end of the block or subprogram, in this case being the main program. Finding the reserved word exception in line 43, the system looks for the desired exception handler, which it finds and executes, then drops out of the bottom of the main program and returns to the operating system.
If there were no handler for the exception, the exception would be propagated to the operating system, and it would issue some sort of nasty message on the standard output device about an unhandled exception leading to program termination.
It should be somewhat obvious to you that if you added another level of subprogram nesting, you could report the error yourself, and possibly recover operation of the program. It is all a matter of program definition.
CAN YOU EXECUTE AN EXCEPTION WITHOUT RAISING IT?
As mentioned before, the section of code at the end of a block or subprogram, following the reserved word exception, is never executed without raising an exception. It can never be executed by dropping into it.
Be sure to compile and execute this program to observe the operation of the exceptions.
EXCEPTIONS CAN OCCUR DURING DECLARATIONS
Example program ------> e_c17_p4.ada
Examine the program named e_c17_p4.ada for an example of an exception that occurs during the declaration part of the program.
When a procedure is called, its declarations are elaborated prior to the logic being executed, as we have stated before. If one of the declarations cannot be properly elaborated, then an error occurs and an exception is raised. Examining the procedure Try_It will reveal an error in lines 8 through 10, where he variable Funny is declared to be of type LIMIT_RANGE with limits of 14 through 33, then it is initialized to the value 8. Since this is out of the allowed range, the exception Constraint_Error will be raised. The executable part of the procedure is not yet ready for use, so the exception handler defined within it cannot be used, and the exception will be propagated to the calling program where it will be handled just as if it occurred in the calling statement, which is line 22 in this case. The exception will therefore be handled by lines 24 through 26 of the procedure Try_To_Fix_It. Note that we never executed the code in the procedure named Try_It.
Be sure to compile and run this program, then study the results.
ADDITIONAL PREDEFINED EXCEPTIONS
You will find that there are actually additional exceptions predefined by your compiler, but these are all defined in additional packages supplied with your compiler. Packages such as Ada.Text_IO, Ada.Sequential_IO, or Ada.Calendar (to be discussed later with tasking), have some number of exceptions defined as a part of their interfaces, but there are only four exceptions predefined as a part of Ada. These were listed and discussed earlier.
A FEW MORE TOPICS CONCERNING EXCEPTIONS
Example program ------> e_c17_p5.ada
The example program named e_c17_p5.ada illustrates a few additional topics about exceptions and illustrates how they are used in a package. This is a very strange program with lots of exception handling examples for your study. You will be left on your own to study the overall operation of this program, but the unique exception handling techniques will be pointed out to you.
The package body contains a section of initialization code in lines 37 through 49 which is composed of nothing but a null statement and several exception handlers. These are only used during initialization of the package since they are not within the executable portion of either of the subprograms. You will notice that the exception named Funny_Add_Error is declared in the package specification so it is visible in the exception handler in line 46, but the exception named Funny_Subtract_Error is not visible there because it is declared within the function. We will see soon however, that even this exception can be propagated to the main program.
When the program is executing, a call to the function Subtract_One raises the exception Funny_Subtract_Error which is handled by the exception handler at the end of the function in line 32. A message is displayed and the same exception is raised by the isolated raise statement in line 34. This statement simply raises the exception that caused the jump to the exception handler in the first place. The isolated raise statement can only be used within an exception handler. The exception is passed on to the calling program, even though it has already been handled here.
Because the exception named Funny_Subtract_Error is not visible to the main program, it cannot handle it by name but even this exception can be handled by an others clause as is done in line 68. After printing a message, the same exception is once again raised in line 70 where it is passed on to the operating system. You will see when you execute this program that the exception is known by name to the operating system. It will give you a nasty message about the unhandled exception and terminate operation.
THE others CLAUSE IN AN EXCEPTION HANDLER
If the others clause is used, it must be the last in the exception handler list and it cannot be combined with any other exceptions such as illustrated in line 64. As in other Ada constructs, two or more exceptions can be "or"ed and use the same exception handler. Numeric_Error was available inn Ada 83, but is considered obsolete in Ada 95. The name Numeric_Error is a synonym for Constraint_Error in Ada 95 to permit legacy code to compile without error.
Note line 59 where an exception is renamed to reduce the length of its name. Any exception can be renamed in a similar fashion.
Be sure to compile and execute this program and spend the time necessary to understand the exception propagation illustrated here.
WHAT IS AN EXCEPTION OCCURRENCE?
Many times it is adequate to simply report that an exception occurred and the general nature of the exception is sufficient to permit a recovery from the exceptional condition. There are times however, that it is necessary to provide additional information about the exception and what caused it. The exception occurrence is available in a predefined Ada 95 package for this purpose. The exception occurrence gives a unique name to one instance of raising an exception and provides hooks to completely analyze where and when the exception occurred. As always in this tutorial, an example program provides the best illustration, so examine the example program named e_c17_p6.ada.
Example program ------> e_c17_p6.ada
In line 2 we with the package named Ada.Exceptions which provides us with the ability to name and use an exception occurrence, and we name one in line 8. The name Except_ID can be used to refer to any single exception at a time, but can be reused to refer to any other exception at a later time. This example program executes a loop with a contrived error in it, a divide by zero. In fact, you may recognize it as a modification of the first example program in this chapter.
When we arrive at the divide by zero condition, the exception Constraint_Error is raised and the exception handler defined in line 23 is found and execution begins. However, because of the occurrence name given immediately following the reserved word when, this particular occurrence of Constraint_Error is given the name Except_ID. Once we have the occurrence name, we can use it to retrieve a formatted message with the name of the exception, even if the exception name in no longer within scope. This is done via a call to Exception_Name which is a part of the Ada.Exceptions package. This is illustrated in line 28 where the returned line of text is simply copied to the monitor. We can also get a more detailed message including a complete call stack listing by calling the function named Exception_Message with the name of the exception occurrence as the only parameter. This result is also copied to the monitor for inspection.
The actual format is implementation dependent, and your compiler may give a very verbose line of text or a very sparse line of text. The compiler used for this compilation emitted a very scarce text.
If we were to correct the error and allow the program to continue, another divide by zero condition would cause the Constraint_Error and the Except_ID exception occurrence could be used to analyze that particular exception. It should be obvious that you could use this to log certain exceptions to a log file for post execution analysis of what caused some particular catastrophic failure. If properly planned, such a log file would have a complete history of system operation prior to failure.
USING THE OCCURRENCE MORE THAN ONCE
In line 34 of the example program, we use the same occurrence name to retrieve and display information about any occurrence of the Storage_Error exception. This indicates that the exception occurrence acts just like a variable and can be used to refer to any exception that occurs during execution of the program. Note that it is very unlikely that the Storage_Error exception will be raised in this program, but is given here only for illustration.
The same name can be used with the others condition to cover any otherwise unhandled exceptions that occur.
Be sure to compile and execute this program with your compiler to see the output format provided for you. There is no standard way to format this data, so your output could look very different from the example output provided in the result of execution. The same information will be provided for you in some meaningful manner.
Return to the Table of Contents